Test analytics
for engineers and agents
The missing test layer for GitHub: regressions, flakes, artifacts, and history in one place.
Any test, any runner
Frontend in Vitest. E2E in Playwright. Backend in Pytest or JUnit.
Flakiness.io brings them into one test layer, with high-quality reporters for the major runners, a JUnit XML bridge for everything else, and a Node.js SDK for custom integrations. Mixed stack, one history.
From failed run to root cause
A failed run carries a lot of signal. Flakiness.io puts the test waterfall, system telemetry, root-cause bins, and Flakiness Query Language in one place.
Engineers and agents can see the failure, understand it, and decide where to look next.
Keep evidence attached
Logs, screenshots, videos, and traces stay attached to the step that produced them. Flakiness.io keeps that evidence in context, with built-in image diffing and Playwright trace viewing in the browser.
See a report with attachmentsDefragment your CI
Shards finish at different times. Environments drift. CI shows the run in pieces. Flakiness.io ingests results as they land, merges shards into one report, and keeps staging and production histories separate. GitHub OIDC keeps authentication tokenless.
GitHub-native access
Flakiness.io follows your GitHub repository permissions. People with access to a repository can see its test analytics; people without can't.
You do not need to recreate users, teams, permissions, or single sign-on (SSO) inside Flakiness.io. Access stays where your engineering team already manages it: GitHub.
Separate regressions from flakes
Every result is tied to the commit and environment it ran on. Flakiness.io can tell whether a failure is new in the PR, already broken on main, or flipping on the same commit. Triage gets faster because the signal is cleaner.
Compact context for coding agents
Raw CI output is expensive context for coding agents. Flakiness.io turns a failed run into a compact record: what failed, whether it regressed, how it behaved on main, and which logs and artifacts matter.
Agents spend fewer tokens gathering context and less time fixing problems.
Claude Code
Codex
CursorWe charge for storage,
not tests or seats
Most platforms charge per test run, per dashboard user, or both. Flakiness.io charges for stored data. Run more tests and let everyone with GitHub access view the results — without per-run or per-seat fees.
Choose your plan
Starts free — no credit card required.
Hobby
- Core test analytics
- 1 GB artifacts storage
- 90 days data retention
- Unlimited test runs
- Unlimited users
- Public projects
- Private projects
- Slack Notifications
- Self-hosting
- Standard support
Startup
- Core test analytics
- 10 GB artifacts storage
- 180 days data retention
- Unlimited test runs
- Unlimited users
- Public projects
- Private projects
- Slack Notifications
- Self-hosting
- Standard support
Growth
- Core test analytics
- 50 GB artifacts storage
- 360 days data retention
- Unlimited test runs
- Unlimited users
- Public projects
- Private projects
- Slack Notifications
- Self-hosting
- Priority support
Enterprise
- Core test analytics
- Custom artifacts storage
- Custom data retention
- Unlimited test runs
- Unlimited users
- Public projects
- Private projects
- Slack Notifications
- Self-hosting
- Priority support
Frequently asked questions
Why is Flakiness.io cheaper than other platforms?
Flakiness.io charges for stored data, not per user or test run. The analytics engine is built on interval unions, which keeps large test histories efficient to pack and process at scale. Pricing follows that architecture: storage drives cost, not execution count.
How does Flakiness.io detect flakiness?
Every test result is tied to the commit and environment it ran on. A test that flips between pass and fail on the same commit is classified as a flake; a test whose outcome changes at a specific commit and stays that way is a regression. This separates noise from real breakage without rerunning anything.
The commit-aware analysis behind this is grounded in research published by Apple engineers.
Can Flakiness.io handle our massive monorepo?
Yes. Flakiness.io is built for large test histories, mixed stacks, and high test volume. We have tested the system on established projects with 500,000+ tests, and it handled the volume without issue.
Can I host Flakiness.io on my own servers?
Yes. A self-hosted deployment is one application container backed by PostgreSQL and an S3-compatible object store. The reporters and the Flakiness CLI work with custom deployments. See the self-hosting docs or contact [email protected] for licensing.
Who develops the test runner integrations?
The Flakiness JSON Report format is open source, so anyone can build a reporter.
The Flakiness.io team maintains the official reporters for the major test runners, the JUnit bridge, and the platform itself. See the test runners documentation for the full list.
More questions? Reach out at [email protected]
Who's behind Flakiness.io
A decade of building dev tools
Flakiness.io is created by Andrey Lushnikov, who spent over a decade at Google and Microsoft building the tools that testing engineers rely on every day: creating Google Puppeteer and co-founding Microsoft Playwright.
After years of dealing with flaky tests and fragmented test reporting, he built the tool he wished existed.